Pharmacometabolomic Assessments of Atenolol and Hydrochlorothiazide Treatment Reveal Novel Drug Response Phenotypes

Abstract

Achieving hypertension (HTN) control and mitigating the adverse health effects associated with HTN continues to be a global challenge. Some individuals respond poorly to current HTN therapies, and mechanisms for response variation remain poorly understood. We used a nontargeted metabolomics approach (gas chromatography time-of-flight/mass spectrometry gas chromatography time-of-flight/mass spectrometry) measuring 489 metabolites to characterize metabolite signatures associated with treatment response to anti-HTN drugs, atenolol (ATEN), and hydrochlorothiazide (HCTZ), in white and black participants with uncomplicated HTN enrolled in the Pharmacogenomic Evaluation of Antihypertensive Responses study. Metabolite profiles were significantly different between races, and metabolite responses associated with home diastolic blood pressure (HDBP) response were identified. Metabolite pathway analyses identified gluconeogenesis, plasmalogen synthesis, and tryptophan metabolism increases in white participants treated with HCTZ (P < 0.05). Furthermore, we developed predictive models from metabolite signatures of HDBP treatment response (P < 1 × 10(-5)). As part of a quantitative systems pharmacology approach, the metabolites identified herein may serve as biomarkers for improving treatment decisions and elucidating mechanisms driving HTN treatment responses.

Figure 1

Histograms of change in HDBP in participants treated with HCTZ or ATEN. Each panel represents the distribution of change in HDBP with columns stratifying groups by treatment of ATEN, HCTZ, or combining all participants. Rows combine all participants or stratify by race. Panels with matching letters represent groups that displayed statistically significant differences in HDBP response (q < 0.1).

Figure 2

Hierarchical clustering of metabolites altered upon treatment of HCTZ (q < 0.2). The boxes on the heat map outline significant modules and are annotated with the metabolites within each module. The x‐axis and y‐axis are metabolites and the shade of color represents the Pearson correlation.

Figure 3

Multivariable modeling of change in HDBP using baseline metabolic signatures. (a) Shows for each model the p value obtained for the discovery (red) and validation (blue) sets. The vertical dashed line represents the Bonferroni corrected P value required for the model results to be considered statistically significant. The x‐axis is the –log₁₀ (P value) so that points to the right represent increasing significance. (b) The R² of each model separated by discovery (red) and validation (blue) sets. The error bars represent the bootstrapped 95% confidence interval.